Prenatal exposure to antidepressants and language competence at age three: results from a large population-based pregnancy cohort in Norway

Authors


Abstract

Objective

To examine the association between maternal use of selective serotonin reuptake inhibitors (SSRI) in pregnancy and language competence in their children at age three taking into account maternal symptoms of anxiety and depression.

Design

Population-based prospective pregnancy cohort study.

Setting

The Norwegian Mother and Child Cohort Study; recruited pregnant women from 1999 through 2008.

Population

45 266 women with 51 748 singleton pregnancies.

Methods

The association between short- or long-term use of SSRI during pregnancy and language competence in the child was investigated using multinomial logistic regression with three outcome categories: long, complicated sentences, fairly complete sentences and language delay.

Main outcome measures

Children's language competence at age three measured by maternal report on a validated language grammar scale.

Results

Women reported use of SSRI in 386 (0.7%) pregnancies. Of these, 161 (42%) reported long-term use. Compared with children whose mothers took no SSRI, using the best language category as the reference, adjusted relative risk ratios (RRR) of having fairly complete sentences were 1.21 (95% CI 0.85–1.72) and 2.28 (1.54–3.38) for short- and long-term SSRI use, respectively. The adjusted RRRs of language delay were 0.86 (0.42–1.76) and 2.30 (1.21–4.37). Symptoms of anxiety and depression in pregnancy were independently related to language delay, adjusted RRR 1.25 (1.03–1.50) and 1.83 (1.40–2.40) for short- and long-term symptoms, respectively.

Conclusions

Prolonged use of SSRI during pregnancy was associated with lower language competence in children by age three independently of depression. Having symptoms of depression throughout pregnancy had an independent effect.

Introduction

The prevalence of depression during pregnancy is estimated at between 7 and 15%.[1, 2] Earlier studies have shown that untreated maternal depression may pose a threat to the mother and the fetus.[3] When medical treatment of pregnant women is necessary, selective serotonin reuptake inhibitors (SSRIs) is the most common treatment.[4]

Limited information is available on the potential effect of prenatal exposure to SSRIs on long-term neurocognitive function in children.[5] Available studies on language development in children exposed to antidepressants during pregnancy have shown no significant association between use of SSRIs and lower language competence.[6, 7] In these studies the population size was small[6] or the language development was examined at an early age, the latest being 19 months.[7]

In Norway, a large prospective population-based pregnancy cohort, the Norwegian Mother and Child Cohort Study (MoBa), has been established.[8] The study is designed to follow-up children's neurodevelopment long-term.

The aim of the present study was to assess the effects of SSRI exposure during pregnancy on language competence in children at 3 years of age while accounting for symptoms of anxiety and depression before, during and after pregnancy.

Methods

This is a population-based prospective pregnancy cohort study based on data from MoBa and the Medical Birth Registry of Norway (MBRN). Self-reported drug use, data on symptoms of anxiety and depression and confounders were obtained prospectively several times during pregnancy. The outcome measure was language competence at age 3 years as reported by the mother. To explore the validity of the exposure data (SSRI use), prescription data from the Norwegian Prescription Database (NorPD) were also used.[9] The three data sources were linked using the unique personal identity number assigned to all individuals living in Norway. If information was available from several data sources, data from MBRN were preferred because of completeness.

Data sources

The Norwegian Mother and Child Cohort Study (MoBa)

MoBa is a prospective population-based pregnancy cohort described in detail elsewhere.[8] The actual sampling frame consists of pregnant women who attended the routine ultrasound appointment around week 17–18 of pregnancy at participating Norwegian hospitals. The final cohort consisted of 90 700 women who gave their written consent to participate (38.5% participation rate), and 108 000 children. The recruitment began in 1999 and ended in 2008. The data collection included three questionnaires during pregnancy, and questionnaires at child age six and 18 months, three, five, seven and 8 years of age.

Medical Birth Registry of Norway (MBRN)

MBRN is a nationwide registry that is based on the compulsory notification of every birth or late abortion from 12 weeks of gestation onwards in Norway.[10] MBRN was used to get information on date of start and end of pregnancy and background characteristics of pregnant women. Estimated gestational age at delivery and thus the start of pregnancy was mostly based on an ultrasound examination in the second trimester (97%) but if such an examination was not carried out, the last menstrual period was used.

Study population

This study is based on data from pregnant women who participated in MoBa (Data file version 7) and from their children, all of whom had reached age 3 years. The age-three questionnaire was returned for 58 410 children. For our analyses we excluded children from pregnancies with multiple fetuses (twins and triplets, n = 1748) and children with malformations and/or chromosomal abnormalities (n = 1555). Pregnancies where the woman did not answer all three pregnancy questionnaires were excluded (n = 3069). In addition we excluded children with missing data on the language measurement (n = 290). Some mothers had more than one pregnancy during the MoBa recruitment period and the study population thus consisted of 45 266 mothers with 51 748 children.

Self-reported SSRI use

The mothers received three questionnaires that included questions regarding drug use in pregnancy which covered: (i) the time period from pregnancy week 0 to week 17–18 of pregnancy; (ii) the time period from week 19 to week 29 of pregnancy, and (iii) the time period from week 30 of pregnancy until the child was 6 months old. Complete questionnaires are available on the Norwegian Institute of Public Health's website (www.fhi.no). A more detailed description of self-reported drug use in MoBa has been described in Skurtveit et al.[9] All SSRIs were studied as one exposure drug group. Antidepressant drugs other than SSRIs were not investigated as the prevalence of such use in the MoBa cohort is very low, 0.1 and 0.2%, respectively, for tricyclic antidepressants and other antidepressants.[4]

Women were defined as users of SSRIs if they had reported use of any SSRI at any time on the 17–18-week questionnaire (pregnancy week 0–18), or the 30-week questionnaire (week 19–29) or the 6-month questionnaire (week 30 until birth). Duration of SSRI medication was classified as: (i) no use, (ii) use in one time period only, and (iii) use in at least two time periods.

The main indication for use of SSRI in Norway is depression but it is also prescribed for anxiety, obsessive-compulsive disorder, bulimia, post-traumatic stress syndrome and cataplexy.

Validity of self-reported SSRI use

The Norwegian Prescription Database, established in 2004, includes data on all prescriptions redeemed by individuals in ambulatory care in Norway. The establishment of this register gave us the opportunity to validate self-reported data from MoBa from pregnancies starting after 2004 with prescription data from the NorPD. We found a redeemed prescription of an SSRI in NorPD for 93% of the women who had reported such use in MoBa. However, among women who had been dispensed an SSRI during pregnancy, only 73% reported that they had used such a drug during pregnancy. The percentages of total agreement for SSRI were substantial (99.5%; kappa 0.69). A more comprehensive validation study of self-reported SSRI use in MoBa has been published recently.[9]

Measure of language competence

Language competence at age 3 years was determined by a language grammar rating scale[11] in the 3-year questionnaire. The mother was asked to choose one category that best described the way her child talked:

  • 1 Talking in long and complicated sentences, such as ‘when I went to the park, I went on the swings’ or ‘I saw a man standing on the corner’.
  • 2 Talking in fairly complete sentences, such as ‘I have a doll’ or ‘Can I go outside?’.
  • 3 Talking in two- to three-word phrases, such as ‘me got ball’ or ‘give doll’.
  • 4 Talking in one-word utterances, such as ‘milk’ or ‘down’.
  • 5-6 Talking but unintelligible/not yet talking.

Categories 5 and 6 were combined in our analyses due to very small numbers. In cases where the mother marked several categories, the child was classified in the most advanced language category. Parental self-report is generally a good measure of early expressive vocabulary, especially for severe language delay.[11, 12] The validity of the language grammar rating scale has been described earlier by Roth et al.[13]

Possible confounders and effect modifiers

We considered many factors that could be associated with the use of SSRI drugs in pregnancy and with language competence in the child.[13] Information regarding age (maternal and paternal) and parity was retrieved from the MBRN. Education level (maternal and paternal), marital status, information about whether the pregnancy was planned, maternal working status, maternal smoking in pregnancy, use of folic acid and pre-pregnancy body mass index (BMI, calculated as weight in kilograms divided by height in metres squared) were retrieved from the first pregnancy questionnaire.

Symptoms of anxiety and depression were assessed both during pregnancy and after birth. In pregnancy the self-assessments were done in week 17–18 and week 30. After birth they were done when the child was 6 months, 18 months and 3 years old. The assessments were done by a screening instrument designed to identify common psychiatric symptoms (e.g. fearfulness, hopelessness, feeling blue) – the 5-item version (SCL-5) of the Hopkins Symptom Checklist (SCL-25).[14] For each item the mother was asked to tick on a scale from 1 to 4 if she was not bothered, a little bothered, quite bothered or very bothered. The SCL-5 scores were calculated as the mean value of the five items to generate a SCL-5 score ranging from 1 to 4. Presence of symptoms of anxiety and depression was defined by having a score >2.0. For example, a woman who answered being ‘a little bothered’ on all five items (assigned a score 2 on each item) would get a SCL-5 score of 2 and accordingly not be classified as having symptoms. In our study, subjects with two or more missing items on SCL-5 were assigned to missing. When only one item was missing (n = 964; 1.9%), it was replaced by the sample mean value for that item. Short-term symptoms in pregnancy were defined as having symptoms either in week 17–18 or in week 30, whereas long-term depression was defined as having symptoms in both periods.

Depression before pregnancy was assessed in pregnancy week 17–18, defined as ever having experienced three or more of the following symptoms simultaneously for a continuous period of 2 weeks or more for no particular reason (death, divorce, etc.): felt depressed or sad; had problems with appetite or had eaten too much; been bothered by feeling weaker or a lack of energy; really blamed oneself and felt worthless; had problems with concentration or making decisions.

Socio-demographic and lifestyle variables and use of other psychotropic drugs (benzodiazepines and/or benzodiazepine-like drugs or opioids) in pregnancy were categorised as presented in Table 1. Effect modifiers such as breastfeeding and reduced hearing in the child were assessed from the six-and/or 18-month questionnaire, and information on prematurity and major malformation was assessed from MBRN.[10]

Table 1. Maternal use of selective serotonin reuptake inhibitors (SSRI) during pregnancy by parental characteristics. Participants in the Norwegian Mother and Child Cohort Study (= 51 748)
 Use of SSRI P e
NoYes, one period onlyYes, at least two periods
n (%)n (%)n (%)
  1. a

    Up to pregnancy week 8.

  2. b

    Benzodiazepines and benzodiazepine-like drugs.

  3. c

    Assessment was done in pregnancy week 17–18.

  4. d

    Symptoms of anxiety and depression were assessed either in pregnancy week 17–18 or 30 (short term) or in both weeks (long term) by the 5-item version of Hopkins Symptom Checklist. A cut-off of 2.0 was used.

  5. e

    Chi-square test.

Maternal formal education in yearsc (n = 50 664)
<128429 (98.9)52 (0.6)40 (0.5)<0.01
127223 (99.1)40 (0.5)28 (0.4)
13–1622 092 (99.4)82 (0.4)62 (0.3)
≥1712 540 (99.4)48 (0.4)28 (0.2)
Paternal formal education in yearsc (= 48 707)
<1216 315 (99.1)88 (0.5)65 (0.4)<0.001
126217 (98.9)40 (0.6)31 (0.5)
13–1614 134 (99.4)51 (0.4)31 (0.2)
≥1711 677 (99.5)33 (0.3)26 (0.2)
Maternal age in years ( =   51 679)
<254597 (99.1)26 (0.6)17 (0.4)0.37
25–2917 137 (99.3)74 (0.4)48 (0.3)
30–3420 435 (99.3)78 (0.4)62 (0.3)
≥359124 (99.1)47 (0.5)34 (0.4)
Paternal age in years ( n =  51 556)
<251898 (98.6)19 (1.0)7 (0.4)<0.001
25–2911 437 (99.3)49 (0.4)37 (0.3)
30–3420 401 (99.4)64 (0.3)58 (0.3)
≥3517 436 (99.1)91 (0.5)59 (0.3)
Planned pregnancyc (= 51 212)
No8532 (98.5)90 (1.0)40 (0.5)<0.001
Yes42 299 (99.4)133 (0.3)118 (0.3)
Maternal smokingc (= 49 660)
No45 489 (99.3)182 (0.4)121 (0.3)<0.001
Yes3796 (98.1)40 (1.0)32 (0.8)
Maternal alcohol intake in pregnancy ( =   51 651)
No25 455 (99.3)101 (0.4)66 (0.3)<0.01
Yes (probably occasionally)23 887 (99.2)110 (0.5)82 (0.3)
Weekly1923 (98.6)14 (0.7)13 (0.7)
Maternal folic acid supplements in early pregnancya (= 51 748)
No13 835 (99.3)54 (0.4)41 (0.3)0.56
Yes37 527 (99.2)171 (0.5)120 (0.3)
Maternal analgesic opioid use in pregnancy ( =   51 748)
No50 485 (99.3)218 (0.4)152 (0.3)<0.001
Yes877 (98.2)7 (0.8)9 (1.0)
Maternal benzodiazepineb use in pregnancy (= 51748)
No51 014 (99.3)204 (0.4)135 (0.3)<0.001
Yes348 (88.1)21 (5.3)26 (6.6)
Parity ( =   51 679)
024 456 (99.1)132 (0.5)90 (0.4)<0.05
117 607 (99.4)54 (0.3)52 (0.3)
≥29230 (99.4)39 (0.4)19 (0.2)
Marital statusc (= 51 522)
Married or living with partner49 802 (99.3)196 (0.4)147 (0.3)<0.001
Single887 (97.0)18 (2.0)9 (1.0)
Other452 (97.6)6 (1.3)5 (1.1)
Maternal BMI, kg/m2c (= 50 627)
<2534 961 (99.3)144 (0.4)96 (0.3)<0.05
25–2910 811 (99.2)45 (0.4)39 (0.4)
30–343309 (99.1)18 (0.5)13 (0.4)
≥351174 (98.6)11 (0.9)6 (0.5)
Maternal depression before pregnancyc (= 50 620)
No47 619 (99.6)122 (0.3)84 (0.2)<0.001
Yes2627 (94.0)98 (3.5)70 (2.5)
Maternal symptoms of anxiety and depression during pregnancyd (= 50 515)
No45 997 (99.6)117 (0.3)79 (0.2)<0.001
Yes, short term3073 (97.6)47 (1.5)30 (1.0)
Yes, long term1066 (91.0)58 (4.9)48 (4.1)
Maternal working statusc (= 51 538)
Working47 313 (99.4)175 (0.4)131 (0.3)<0.001
Not working2832 (99.0)18 (0.6)11 (0.4)
Disability pensioner335 (90.3)21 (5.7)15 (4.0)
Other674 (99.8)10 (1.5)3 (0.4)

Analysis strategy and statistical analysis

As the language competence outcome has several categories, ranging from long, complicated sentences to unintelligible utterances/not yet talking, we used multinomial logistic regression after collapsing the three worse outcomes due to small numbers. The three outcome categories were: long, complicated sentences (talking in long and complicated sentences); fairly complicated sentences (talking in fairly complete sentences); language delay (talking in two- to three-word phrases, talking in one-word utterances, talking but unintelligible/not yet talking). The language delay category comprises both moderate and severe language delay. We estimated standard errors using the clustered sandwich estimator, allowing for clustering of multiple pregnancies within the same woman. The outcome measure was the relative risk ratio (RRR) using the best language category as the reference. The interpretation is similar to odds ratios in logistic regression.

To disentangle the effect of SSRI use and symptoms of anxiety and depression during pregnancy we adjusted for depression before pregnancy and depression during pregnancy, in addition to other possible confounders. Moreover, we did stratified analyses according to whether the mother reported symptoms of anxiety or depression when the child was 6 months, 18 months and 3 years because depression after birth could be a possible effect modifier. In addition, stratified analyses on symptoms of maternal depression at 3 years is an attempt to account for under- or overestimation in the maternal report of the child's language competence among women with depression and anxiety.

Stratified analyses were also performed on child gender as a possible effect modifier.

To check for possible selection bias due to no response to the 3-year questionnaire, we compared the prevalence of use of SSRIs during pregnancy among those who responded to the 3-year questionnaire with the use of SSRI among non-responders to the questionnaire. The prevalence of use of SSRIs was 0.7% in the study population and 0.8% among non-responders.

Statistical analyses were conducted using SPSS for Windows, 17.0.1 (SPSS Inc., Chicago, IL, USA) and STATA 12.

Results

In all, 45 266 women with 51 748 pregnancies were included in this study. The women (n = 373) reported use of SSRIs in 386 pregnancies (0.7%). Of these, 161 (42%) reported use of SSRI during at least two time periods during pregnancy. Of all children (n = 51 748) aged 3 years, 39 841 (77%) were rated as talking in long and complicated sentences, 9847 (19%) as talking in fairly complete sentences, 1727 (3.3%) as talking in two- to three-word phrases, 192 (0.4%) as talking in one-word utterances, and 141 (0.3%) as not yet talking, or talking unintelligibly.

Table 1 shows maternal use of SSRIs in pregnancy by parental characteristics. Use of SSRIs was more common in the groups where the mother and father had a lower education. It was more common in single mothers, mothers who reported smoking during pregnancy, and unplanned pregnancies. Among the potential confounders, parental education, maternal working status and marital status in pregnancy, parity, smoking and BMI were strongly associated with both exposure and outcome (Tables 1 and 2).

Table 2. Language competence in children by parental characteristics. Participants in the Norwegian Mother and Child Cohort Study (= 51 748)
 Long, complicated sentencesFairly complete sentencesTwo- to three-word phrasesOne-word utterancesUnintelligible utterances + not yet talking P e
n (%)n (%)n (%)n (%)n (%)
  1. a

    Up to pregnancy week 8.

  2. b

    Benzodiazepines and benzodiazepine-like drugs.

  3. c

    Assessment was done in pregnancy week 17–18.

  4. d

    Symptoms of anxiety and depression were assessed either in pregnancy week 17–18 or 30 (short term) or in both weeks (long term) by the 5-item version of Hopkins Symptom Checklist. A cut-off of 2.0 was used.

  5. e

    Chi-square test.

Maternal formal education in years c ( =   50 664)
<125890 (69.1)2032 (23.8)500 (5.9)63 (0.7)36 (0.4)<0.001
125498 (75.4)1459 (20.0)273 (3.7)30 (0.4)31 (0.4)
13–1617 435 (78.4)4098 (18.4)600 (2.7)58 (0.3)45 (0.2)
≥1710 199 (80.8)2043 (16.2)312 (2.5)37 (0.3)25 (0.2)
Paternal formal education in yearsc (= 48 707)
<1211 995 (72.8)3568 (21.7)753 (4.6)93(0.6)58 (0.4)<0.001
124853 (77.2)1197 (19.0)197 (3.1)20 (0.3)21 (0.3)
13–1611 275 (79.3)2510 (17.7)370 (2.6)30 (0.2)31 (0.2)
≥179450 (80.5)1960 (16.7)266 (2.3)34 (0.3)26 (0.2)
Maternal age in years ( =   51 679)
<253518 (75.8)887 (19.1)199 (4.3)20 (0.4)16 (0.3)<0.001
25–2913 598 (78.8)3058 (17.7)511 (3.0)57 (0.3)35 (0.2)
30–3415 823 (76.9)3939 (19.1)678 (3.3)79 (0.4)56 (0.3)
≥356846 (74.4)1952 (21.2)338 (3.7)35 (0.4)34 (0.4)
Paternal age in years ( =   51 556)
<251479 (76.9)359 (18.7)75 (3.9)5 (0.3)6 (0.3)<0.001
25–299118 (79.1)2007 (17.4)330 (2.9)39 (0.3)29 (0.3)
30–3416 059 (78.2)3722 (18.1)62 (3.1)71 (0.3)45 (0.2)
≥3513 040 (74.1)3726 (21.2)684 (3.9)75 (0.4)61 (0.3)
Planned pregnancyc (= 51 212)
No6494 (75.0)1741 (20.1)362 (4.2)36 (0.4)29 (0.3)<0.001
Yes32 954 (77.4)7986 (18.8)1348 (3.2)151 (0.4)111 (0.3)
Maternal smokingc (= 49 546)
No35 602 (77.7)8506 (18.6)1397 (3.1)166 (0.4)121 (0.3)<0.001
Yes2673 (71.3)854 (22.6)207 (5.5)10 (0.3)10 (0.3)
Maternal alcohol intake in pregnancy ( =   51 651)
No19 937 (77.8)4711 (18.4)801 (3.1)104 (0.4)69 (0.3)<0.001
Yes (probably occasionally)18 329 (76.1)4750 (19.7)854 (3.5)76 (0.3)70 (0.3)
Weekly1510 (77.4)366 (18.8)67 (3.4)5 (0.3)2 (0.1)
Maternal folic acid supplements in early pregnancya (= 51 748)
No10 163 (73.0)3012 (21.6)624 (4.5)78 (0.6)53 (0.4)<0.001
Yes29 678 (78.5)6835 (18.1)1103 (2.9)114 (0.3)88 (0.2)
Maternal analgesic opioid use in pregnancy ( n =  51 748)
No39 178 (77.0)9654 (19.0)1699 (3.3)189 (0.4)135 (0.3)<0.05
Yes663 (74.2)193 (21.6)28 (3.1)3 (0.3)6 (0.7)
Maternal benzodiazepineb use in pregnancy (= 51 748)
No39 551 (77.0)9767 (19.0)1708 (3.3)188 (0.4)139 (0.3)0.07
Yes290 (73.4)80 (20.3)19 (4.8)4 (1.0)2 (0.5)
Parity ( =   51 679)
019 880 (80.6)3995 (16.2)674 (2.7)72 (0.3)57 (0.2)<0.001
113 288 (75.0)3661 (20.7)635 (3.6)78 (0.4)51 (0.3)
≥26617 (71.2)2180 (23.5)417 (4.5)41 (0.4)33 (0.4)
Marital statusc (= 51 522)
Married or living with partner38 667 (77.1)9525 (19.0)1629 (3.2)186 (0.4)138 (0.3)<0.001
Single657 (71.9)190 (20.8)60 (6.6)4 (0.4)3 (0.3)
Other351 (75.8)87 (18.8)24 (5.2)1 (0.2)0 (0)
Maternal BMI, kg/m2c (= 50 627)
<2527 583 (78.4)6327 (18.0)1090 (3.1)116 (0.3)85 (0.2)<0.001
25–298254 (75.8)2187 (20.1)381 (3.5)41 (0.4)32 (0.3)
30–342400 (71.9)772 (23.1)137 (4.1)18 (0.5)13 (0.4)
≥35784 (65.8)310 (26.0)82 (6.9)10 (0.8)5 (0.4)
Maternal depression before pregnancyc (= 50 620)
No36 911 (77.2)9057 (18.9)1560 (3.3)170 (0.4)127 (0.3)<0.05
Yes2115 (75.7)536 (19.2)120 (4.3)14 (0.5)10 (0.4)
Maternal symptoms of anxiety and depression during pregnancyd (= 50 515)
No35 834 (77.6)8646 (18.7)1452 (3.1)152 (0.3)109 (0.2)<0.001
Yes, short tem2336 (74.2)651 (20.7)132 (4.2)16 (0.5)15 (0.5)
Yes, long term816 (69.6)258 (22.0)78 (6.7)11 (0.9)9 (0.8)
Maternal working statusc (= 51 538)
Working36 976 (77.6)8889 (18.7)1479 (3.1)157 (0.3)118 (0.2)<0.001
Not working1955 (68.3)688 (24.0)171 (6.0)29 (1.0)18 (0.6)
Disability pensioner242 (65.2)93 (25.1)30 (8.1)4 (1.1)2 (0.5)
Other527 (76.7)127 (18.5)31 (4.5)1 (0.1)1 (0.1)

Table 3 shows the proportions of children with different language development according to maternal use of SSRIs and reported symptoms of anxiety and depression during pregnancy. We observed a shift in the distribution towards lower language competence, the longer the mothers had used SSRIs during pregnancy. Only 57% of the children exposed to SSRIs during at least two periods fell in the best language competence category, compared with 77% in children who were not exposed. Moderate delay, that is, use of two- to three-word phrases, was also more common among children exposed to SSRI during at least two periods (7%) than among unexposed children (3%). It was not possible to analyse severe language delay as there were very few children in the categories with one-word utterances or more severe delay in the SSRI groups. A shift in the distribution towards lower language competence was also observed for children whose mothers reported symptoms of anxiety and depression during pregnancy. The shift from the best language competence category was less pronounced than for SSRIs. We saw an association between moderate delay (two- to three-word phrases) and symptoms, similar to the association between moderate delay and SSRIs.

Table 3. Number and proportion (%) of children in the different language categories as reported by the mother in the 3-year questionnaire by selective serotonin reuptake inhibitor (SSRI) use in pregnancy (= 51 748) and by mother's symptoms of anxiety and depression in pregnancy (= 50 515)
 Outcome - language competence
Long, complicated sentencesFairly complete sentencesTwo- to three-word phrasesOne-word utterancesUnintelligible utterances + not yet talking
  1. Fisher's exact test for the first part (use of SSRI) of the table and from chi-square test for the second part (symptoms of anxiety/depression) gave P-values <0.001.

  2. a

    Symptoms of anxiety and depression were assessed either in pregnancy week 17–18 or 30 (short term) or in both weeks (long term) by the 5-item version of Hopkins Symptom Checklist. A cut-off of 2.0 was used.

Use of SSRI
No39 590 (77.1)9736 (19.0)1707 (3.3)190 (0.4)139 (0.3)
Yes251 (65.0)111 (28.8)20 (5.2)2 (0.5)2 (0.5)
Use in one time period only159 (70.7)56 (24.9)8 (3.6)1 (0.4)1 (0.4)
Use in at least two time periods92 (57.1)55 (34.2)12 (7.5)1 (0.6)1 (0.6)
Symptoms of anxiety/depression a
No35 834 (77.6)8646 (18.7)1452 (3.1)152 (0.3)109 (0.2)
Yes3152 (72.9)909 (21.0)210 (4.9)27 (0.6)24 (0.6)
Short term2336 (74.2)651 (20.7)132 (4.2)16 (0.5)15 (0.5)
Long term816 (69.6)258 (22.0)78 (6.7)11 (0.9)9 (0.8)

Results from the multinomial logistic regression are shown in Table 4. Adjusting for maternal depression before pregnancy, the duration of symptoms of anxiety and depression during pregnancy and other confounders did not change the risk estimate substantially, resulting in an adjusted RRR of 2.28 for being in the group with fairly complete sentences to being in the group with the best language competence after long-term SSRI exposure (Table 4). The risk of language delay was also increased after long-term exposure, RRR = 2.30 for language delay versus long, complicated sentences. Maternal depression before pregnancy had no impact on language competence, whereas symptoms of anxiety and depression during pregnancy were independently related to language delay. Additional adjustment for use of analgesic opioids and benzodiazepines/benzodiazepine-like drugs had a negligible impact on measured associations (data not shown).

Table 4. Relative risk ratios (RRR) from multinomial logistic regression of having a child with lower language competence according to selective serotonin reuptake inhibitor (SSRI) use, maternal depression before pregnancy, and symptoms of anxiety and depression (short and long term) in pregnancy (= 43 323)
 Unadjusted RRRbModel 1 adjusted RRR (95% CI)Model 2 adjusted RRR (95% CI)Model 3 adjusted RRR (95% CI)
  1. Model 1: Independent variables: SSRI use and maternal depression before pregnancy.

  2. Model 2: Independent variables: SSRI use, maternal depression before pregnancy and symptoms of anxiety/depression in pregnancy.

  3. Model 3: Independent variables: same as model 2 plus maternal work situation, parity, marital status, smoking, BMI and maternal and paternal education.

  4. a

    Symptoms of anxiety and depression were assessed either in pregnancy week 17–18 or 30 (short term) or in both weeks (long term) by the 5-item version of Hopkins Symptom Checklist. A cut-off of 2.0 was used.

  5. b

    The unadjusted analysis was restricted to the same study sample as in model 3.

  6. Significant results are shown in bold text.

Fairly complete sentences vs long, complicated sentences
Use of SSRI
No (reference)    
Use in one time period only1.29 (0.92–1.82)1.29(0.92–1.83)1.23 (0.87–1.73)1.21 (0.85–1.72)
Use in at least two time periods2.41 (1.643.53)2.40 (1.633.54)2.30 (1.553.40)2.28 (1.543.38)
Maternal depression before pregnancy
No (reference)    
Yes1.03 (0.93–1.15)1.00 (0.90–1.12)0.97 (0.87–1.09)0.99 (0.88–1.10)
Symptoms of anxiety/depression in pregnancya
No (reference)    
Short term1.12 (1.011.24) 1.11 (1.011.23)1.06 (0.96–1.18)
Long term1.24 (1.061.46) 1.19 (1.011.41)1.10 (0.93–1.30)
Language delay vs long, complicated sentences
Use of SSRI
No (reference)    
Use in one time period only1.36 (0.69–2.66)1.19 (0.60–2.37)0.92 (0.45–1.88)0.86 (0.42–1.76)
Use in at least two time periods3.44 (1.866.35)3.04 (1.635.68)2.40 (1.274.53)2.30 (1.214.37)
Maternal depression before pregnancy
No (reference)    
Yes1.39 (1.151.69)1.34 (1.091.63)1.17 (0.95–1.44)1.19 (0.96–1.46)
Symptoms of anxiety/depression in pregnancya
No (reference)    
Short term1.45 (1.201.75) 1.41 (1.171.70)1.25 (1.031.50)
Long term2.45 (1.923.14) 2.27 (1.752.30)1.83 (1.402.40)

The main results (Table 4) did not change with the five alternative analytical approaches: (i) excluding children with reduced hearing; (ii) excluding children with no word production and/or with unintelligible utterances; (iii) excluding children with a birthweight of less than 2500 g; (iv) excluding children born before pregnancy week 37; (v) confirming that the association between exposure and outcome was in the same direction for both genders, for different strata on symptoms of anxiety and depression after pregnancy at 6 months, 18 months and 3 years. We found a significant interaction between SSRI use and folic acid use suggesting a stronger effect of SSRI among users of folic acid (= 0.02).

Discussion

Main findings

In this large population-based pregnancy cohort study we found that prolonged prenatal SSRI exposure was associated with a delay in language competence in children age 3 years independently of maternal symptoms of anxiety and depression before and during pregnancy. Maternal depression after pregnancy did not seem to affect the result. Children of mothers who reported symptoms of anxiety and depression throughout pregnancy also had an increased risk of delayed language competence at age 3 years.

Strengths and limitation of the study

A major strength of our study was the size of the cohort and that the risk of recall bias was minimised by the prospective design. Norway is a country where high-quality and homogeneous universal care during pregnancy and child care is provided to everyone free of charge. Women responded to detailed questions regarding use of prescription drugs and a number of other health-related and socio-demographic factors. This enabled several important potential confounders to be controlled, including duration of depressive symptoms and concomitant use of psychotropic drugs.

A validation of self-reported use of SSRIs showed that 27% of mothers with dispensed drugs during pregnancy did not report drug use on the questionnaire. We do not know whether these women under-reported drug use in the MoBa questionnaire or whether they accurately reported no use, that is, they were dispensed the drug at the pharmacy but did not take the medication. However, under-reporting probably has little impact on effect estimates when the prevalence of exposure is low. The specificity, that is, the proportion of unexposed women correctly classified as unexposed, is more important[9, 15], as false positives would dilute the association when the exposure prevalence is low. In the validation study, only 7% of women who reported use did not have a prescription in NorPD during pregnancy or during the period from 90 days before pregnancy.

Unfortunately, information about SSRI dosage is not available in the MoBa data. However, we did have good estimates of duration of treatment. Due to small numbers, we did not analyse the different SSRIs but collapsed all the different SSRI substances into one group. It is, however, possible that different SSRIs have different effects.

To examine the potential for selection bias related to participation in MoBa, exposure –outcome associations in the MoBa cohort have previously been compared with associations in the MBRN and no evidence of selection bias has been found.[16, 17] We also compared the prevalence of use of SSRIs during pregnancy between responders and non-responders to the 3-year questionnaire and found the prevalence to be very similar.

The SCL-25 has been shown to be an acceptable screen for symptoms of anxiety and depression.[18] In this study, symptoms of anxiety and depression were measured by self-assessment with the SCL-5, a short version of the SCL-25. A validity study of Strand et al.[14] suggested that SCL-5 performed as well as the full 25-item version, at least when considering depression. However, SCL-5 is a soft variable probably hampered by a higher degree of measurement error than exposure to SSRI, and a score above 2 is not a clinical diagnosis. Thus, the degree of under-estimation is potentially a greater problem for symptoms of depression than for use of SSRI.

Interpretation

Our results showed an association between prolonged prenatal exposure to SSRIs and a shift in the distribution of language development towards lower competence in children. The study of prenatal SSRI exposure is complicated by possible effects of the underlying disease. In our study the association between prenatal exposure to SSRIs and lower language competence was still present even after we took into account maternal depression before, during and after pregnancy.

Our results suggest that human fetal brain development is susceptible to SSRI exposure. However, we did not study severe language delay and thus the results do not indicate that prenatal SSRI exposure was associated with a clinical delay in language development. Children with moderate language delay may later catch up with their peers[19] and as we did not study severe language delay, our results may represent less of a long-term impact.

Limited information is available on the effects of maternal SSRI exposure in pregnancy on language competence in children. No difference in mean language scores was found in a study of one- to 6-year-old children, comparing children born to mothers who had taken fluoxetine during pregnancy and children of mothers who were not depressed during pregnancy.[6] A recent study of Pedersen et al.[7] reported a lack of significant association between SSRI exposure and language development in 19-month-old children. These differing results from our findings can possibly be explained by differences in populations or sample sizes, use of different outcome measures, differences in child age at assessment and different data collecting methods (questionnaires compared with psychological tests).

There is also limited information in the literature on neurodevelopment past the neonatal period in children after maternal prenatal depression.[3, 5, 20] An earlier study showed that severity of maternal depression in pregnancy and at testing predicted child behaviour but not cognitive outcomes such as verbal IQ.[21] In contrast, we observed an effect of symptoms of anxiety and depression in pregnancy on the risk of language delay. The effect was stronger for long-term than shorter-term symptoms. In another study the number of maternal depressive episodes after delivery adversely affected child language development.[22]

Mothers’ perception of the child's language development may be impaired by their own psychiatric issues. Stratified analyses according to whether the mother reported symptoms of anxiety or depression when the child was 3 years of age were performed and we found an evident association between SSRI use in pregnancy and language development in both groups of women.

Serotonin is a signalling molecule with a widespread distribution in the central nervous system, thus influencing almost every sphere of the mammalian physiology including cognition, learning and memory.[23] In addition, increasing evidence points to serotonin as a crucial molecule for the modulation of neurodevelopmental processes. Studies in animal models have indicated that serotonin is a key modulator of neuronal cell proliferation, migration and brain wiring during fetal development.[24] SSRI treatments in early life in rats have been shown to have long-lasting effects on both the structure and the function of the somatosensory system.[25, 26] It is not known whether these effects of SSRIs are paralleled in humans, but the findings may indicate that there is a biological explanation for the main finding of our study.

Conclusion

In this large prospective pregnancy cohort in Norway, use of SSRIs during extended periods in pregnancy was associated with risk of the child having lower language competence at age 3 years, independent of symptoms of depression. Although there was a shift in the distribution, very few children could be classified as having clinically impaired language even after long-term prenatal exposure to SSRI. Furthermore, maternal depression was independently associated with language delay. None of these findings should be used as an argument not to treat pregnant women for depression when such treatment is needed.

Disclosure of interests

The authors declare that they have no conflicts of interest.

Contribution to authorship

Svetlana Skurtveit initiated the study and prepared the files for statistical analysis, performed the statistical analysis and drafted the paper. Randi Selmer participated in the study development and statistical analysis, discussion of the paper and the writing of the paper. Sonia Hernandez-Diaz and Christine Roth participated in the study development, discussion of the paper and the writing of the paper. Marte Handal initiated the study and was project manager for the study, and participated in statistical analysis and manuscript preparation. All authors contributed to revising the manuscript and approved the final version to be published.

Details of ethics approval

This MoBa study was approved by the Regional Committee for Medical and Health Research Ethics South- East (s-08176a, April 10th 2008) and the record linkage was approved by The Norwegian Data Protection Authority (08/00854-1/IUR), Jun 25th 2008). Informed consent was obtained from mothers before inclusion in MoBa.

Funding

This study was funded by the Norwegian Institute of Public Health. The funding source had no role in the study design, statistical analysis and interpretation of data, or in the decision to submit the paper for publication.

Acknowledgements

The Norwegian Mother and Child Cohort Study is supported by the Norwegian Ministry of Health and the Ministry of Education and Research, NIH/NIEHS (Contract no NO-ES-75558), NIH/NINDS (grant no.1 UO1 NS 047537-01), and the Norwegian Research Council/FUGE (grant no. 151918/S10). We are grateful to all the participating families in Norway who have taken part in this ongoing cohort study.

Ancillary